Altitudinal patterns in breeding bird species richness and density in relation to climate, habitat heterogeneity, and migration influence in a temperate montane forest (South Korea)

Altitudinal patterns in the population ecology of mountain bird species are useful for predicting species occurrence and behavior. Numerous hypotheses about the complex interactions among environmental factors have been proposed; however, these still remain controversial. This study aimed to identify the altitudinal patterns in breeding bird species richness or density and to test the hypotheses that climate, habitat heterogeneity (horizontal and vertical), and heterospecific attraction in a temperate forest, South Korea. We conducted a field survey of 142 plots at altitudes between 200 and 1,400 m a.s.l in the breeding season. A total of 2,771 individuals from 53 breeding bird species were recorded. Altitudinal patterns of species richness and density showed a hump-shaped pattern, indicating that the highest richness and density could be observed at moderate altitudes. Models constructed with 13 combinations of six variables demonstrated that species richness was positively correlated with vertical and horizontal habitat heterogeneity. Density was positively correlated with vertical, but not horizontal habitat heterogeneity, and negatively correlated with migratory bird ratio. No significant relationships were found between spring temperature and species richness or density. Therefore, the observed patterns in species richness support the hypothesis that habitat heterogeneity, rather than climate, is the main driver of species richness. Also, neither habitat heterogeneity nor climate hypotheses fully explains the observed patterns in density. However, vertical habitat heterogeneity does likely help explain observed patterns in density. The heterospecific attraction hypothesis did not apply to the distribution of birds along the altitudinal gradient. Appropriate management of vertical habitat heterogeneity, such as vegetation cover, should be maintained for the conservation of bird diversity in this area.

135 from the National Center for Environmental Prediction (NCEP) Final (FNL) Operational Global 136 Analysis data. Using these data, climate simulation with WRF was executed for April, May, and 137 June 2015 at time intervals (Δt) of 180 s. Since the NCEP input data resolution of 1° is very 138 coarse for regional or local climate simulations, the domains in this study were downscaled into 139 two-way quadruples of 27, 9, 3, and 1 km with 31 vertical levels in WRF. Simulation outputs 140 were produced every hour with a cumulus parameterization scheme by Kain 148 To quantify vertical habitat heterogeneity, we surveyed the vertical coverage of 149 vegetation at each sampling plot within 5-m radii. Within these circles, we classified vertical 150 layers into understory (< 2 m), midstory (2-10 m), and overstory (> 10 m) vegetation. Coverage 151 was classified into the following four categories: 0 (0 % coverage), 1 (1-33 % coverage), 2 (34-152 66 % coverage), and 3 (67-100 % coverage) ( 165 To identify migration influence, we simply used the migratory bird ratio, which was 166 calculated based on the ratio of the total number of species or individuals and the number of 167 migratory species or individuals in each plot (Helle & Fuller, 1988;Newton & Dale, 1996). All 168 birds detected were classified as residents or summer migrants. Migrants were defined as 169 wintering in the tropical region of Southeast Asia and migrating to the study area for breeding 170 purposes. Twenty-three species were identified as summer migrants and 30 species were defined 171 as residents (Table S2).
Pearson's correlation analysis of nine environmental variables showed that spring 214 temperature and relative humidity were highly correlated (r = -0.951; Table S3). Elevation 215 showed strong correlations with spring temperature and relative humidity (r = -0.977, r = 0.938, 216 respectively; Table S3). Although migratory ratio of species and individuals were correlated (r = 217 0.851; Table S3), these were not included in the same model. Therefore, elevation and relative 218 humidity variables were eliminated from the curve estimation and model construction.

219
In the best-fit curve estimation between species richness, density, and environmental 220 variables, species richness showed significant correlations with spring temperature (R 2 = 0.08, p 221 = 0.003; Fig. 4a) and migratory bird ratio (R 2 = 0.11, p < 0.001; Fig. 4b), and were represented 222 by hump-shaped curves. No relationships were observed between species richness and coverage 223 of understory vegetation, midstory vegetation, or habitat diversity (Fig. 4c, d, and f). Species 224 richness and coverage of overstory vegetation showed a significant positive correlation (R 2 = 225 0.14, p < 0.001; Fig. 4e). Moreover, density showed a significant correlation with spring 226 temperature in a hump-shaped pattern (R 2 = 0.11, p < 0.001; Fig. 5a). A decreasing pattern was 227 observed between density and migratory bird ratio (R 2 = 0.07, p = 0.006; Fig. 5b), and coverage 228 of under-and overstory vegetation represented a monotonically increasing pattern with 229 increasing density (R 2 = 0.03, p = 0.027; R 2 = 0.40, p = 0.017; Fig. 5c and e). Other variables, 230 including coverage of midstory vegetation and habitat diversity, did not show any significant 231 correlations. The set of candidate models with 13 combinations of six variables showed six models 235 supported for species richness (Table 1). The best predictors of species richness were overstory 236 vegetation, midstory vegetation, understory vegetation, habitat diversity, and migratory bird ratio 237 (w i = 0.364). Vertical coverage variables were included in all supported species richness models. 238 A model including habitat diversity was 2.2 times more likely to explain species richness better 239 than models excluding it (w i 0.364 vs. 0.164; Table 1). The Akaike weight was 1.8 times higher 240 the inclusion of migratory bird ratio than when these parameters were excluded (w i 0.364 vs. 241 0.197, Table 1). When spring temperature was excluded in the species model, the Akaike weight 242 was 2.4 times higher (w i 0.364 vs. 0.149, Table 1). Therefore, we regarded spring temperature as 243 an uninformative parameter and excluded it in the next model and constructed seven models 244 again ( Table 2). As a results of seven models, the best supported model was full model (w i 0.488, 245 Table 2). The Akaike weight was 1.8 and 2.2 times higher the inclusion of migratory bird ratio 246 and habitat diversity, respectively, than these parameters were eliminated from model (w i 0.488 247 vs w i 0.264, w i 0.488 vs w i 0.219, Table 2) 248 The results of model selection for predicting density showed three supported models 249 (Table 3). The best model for predicting density included overstory vegetation, midstory 250 vegetation, understory vegetation, habitat diversity, and migratory bird ratio (individuals) (w i = 251 0.342, Table 3). Vertical coverage variables and migratory bird ratio were included in all 252 supported models. When habitat diversity was included in the density model, the Akaike weight 253 was 1.13 times higher than when habitat diversity was eliminated from the model (w i 0.342 vs. 254 0.303; Table 3), and 1.07 times higher in the absence of spring temperature (w i 0.342 vs. 0.321; 255 Table 3). 256 Multimodel-averaged parameter estimates of species richness, including the three 257 supported models, showed positive correlations with overstory vegetation, understory vegetation, 258 and habitat diversity (p < 0.001, p = 0.025, p = 0.040, respectively; Table 4). Density including 259 the three supported models showed positive correlations with overstory vegetation and 260 understory vegetation (p < 0.001, p < 0.001; Table 4) and a negative correlation with migratory 261 bird ratio (p < 0.001; Table 4).   (Table 4), and species richness showed a 294 significant positive relationship with density (Fig. S1). Further, the present study demonstrated 295 that species richness was affected by horizontal habitat diversity, but density was not (Table 4). 296 High habitat diversity can increase species richness due to niche partitioning and providing 297 habitat edges (Best, Whitmore & Booth, 1990), but high habitat diversity does not necessarily 298 indicate high habitat quality with ample food resources. Therefore, the lack of a relationship 299 between density and habitat diversity in this study might be because density increased with 300 productivity and habitat quality (Hurlbert, 2004; Goetz et al., 2007). According to the habitat 301 heterogeneity hypothesis (MacArthur & MacArthur, 1961; Pan et al., 2016), greater structural 302 complexity in vegetation and more habitat types likely contributed to species richness in the 303 present study. However, a larger number of habitat types did not influence the density. 304 We observed a negative relationship between density and migratory bird ratio, and no 305 relationship was observed between species richness and migratory bird ratio ( Fig. 5b and  , we predicted that the migratory bird ratio would have a 308 positive effect on migrant species richness and density, and that the migratory bird ratio would 309 increase with resident species richness and density. However, in the present study, a reduction in 310 the migratory bird ratio led to an increase in density. Additionally, migrant species richness and 311 density showed an increasing pattern along the altitude gradient, whereas resident species 312 richness and density showed a mid-peak pattern along the altitude gradient (Fig. S2). It is 313 unlikely that the migrants could choose a mid-elevation with higher vegetation coverage than the 314 residents could (Fig. S2). Migrant species and individuals did not positively influence species 315 richness and density, and they were not attracted to resident species. Therefore, the 316 heterospecific attraction hypothesis was not applicable along the altitude gradient surveyed in the 317 present study.

318
No relationships were found between species richness or density and climatic factors 319 (Table 4) and a decreasing pattern of spring temperature along the altitudinal gradient was 320 identified (Fig. S2). Numerous studies have shown a positive relationship between temperature 321 and diversity (McCain, 2009). However, significantly stronger relationships between temperature 322 and diversity can be found in humid mountain habitats than in dry mountain habitats (McCain, 323 2009). Furthermore, a negative relationship between density and climatic factors was found in a 324 study conducted in Asia (Pan et al., 2016). Despite numerous studies on this phenomenon, the 325 pattern has not been adequately explained (Currie et al., 2004;Rahbek et al., 2007). Most studies 326 used the average annual temperature from the WorldClim database and conducted bird surveys 327 across all seasons using considerably larger datasets that have constrained accuracy due to the 328 sampling effort involved (Lomolino, 2001;Ding et al., 2005). However, in the present study, we 329 used spring temperature values derived for micro-scale studies, and focused on breeding bird 330 survey on a local scale in a short period in mixed and deciduous forest areas; this approach may 331 have led to the variation in the findings. Another possible explanation is that birds are restricted 332 more by habitat quality for chick rearing than by temperature during the breeding season.

333
A single variable analysis showed no significant relationships between species richness 334 and understory vegetation or habitat diversity (Fig. 4c and f); however, a significant relationship 335 was observed in the modeling approach. Additionally, no differences in density were observed 336 either in the single variable or in the modeling approach. Null hypothesis testing, similar to a 337 simple linear correlation, has been used in many ecological studies and is currently being used in 338 many areas. However, almost ecological phenomena has been often represented by nonlinear and 339 multiple interaction among variables (Landis et al., 2013). For example, species should live at 340 the proper temperature for the optimal thermal fitness during breeding season. But if there is no 341 proper nesting resources, food and shelter, the species should choose a different habitat even the 342 proper temperature for breeding. Consequentially, each variables does not affect the dependent 343 variable, but multiple interactions of the variables. Therefore, alternative modeling approach is 344 required for ecological studies and considered to be a more reliable method that avoiding 345 uninformative, logical deficiencies and common misinterpretations of null hypothesis testing 346 (Anderson, Burnham & Thompson, 2000;Mönkkönen, Forsman & Bokma, 2006). In order to 347 understand the complex ecological phenomena, the use of multimodels is more reasonable and 348 needs more efforts to clarify the relationship of the causative variables. Trends in species richness showed hump-shaped patterns along altitudinal gradients and 352 were related to vertical vegetation coverage and horizontal habitat diversity. In addition, trends 353 in density also showed hump-shaped patterns, with density related to vertical vegetation 354 coverage and migratory bird ratio, but not to habitat diversity. No significant relationships were 355 found between spring temperature and species richness or density. The results on species 356 richness support the habitat heterogeneity hypothesis rather than the climate hypothesis, whereas 357 those of species density do not support fully either hypothesis, and they were related to species 358 richness and vertical vegetation coverage. The heterospecific attraction hypothesis was not 359 applicable to the distribution of birds along the altitudinal gradient studied. Taken together, our 360 findings indicate that management of vegetation cover would be an appropriate strategy for avian 361 conservation in this region. To achieve a better understanding of the specific reasons for the 362 distribution of birds along altitudinal gradients, further studies on the interactions among species 363 related to niche and competition are required.